Computer vision clean-up (archive)

It’s very extremely difficult to remove in landscapes. Not only does it climb things, but it makes these little potato-y root tubers and if you miss any when digging it out, a new plant grows back. You have to destroy the whole area looking for them. But yeah I guess it would be controlled in a pot.

Asparagus setacea is a mess, Barbara is right. I see every other species of Asparagus, horsetails, fennel, yarrow, some conifers, even a Coleonema…

4 Likes

“People keep that nasty weed as a houseplant?!”

Apparently so. Lots of photos show it in pots. At least those are easy to classify.

1 Like

I was surprised to see the ones in pots are correctly identified much more often than the ones in the ground!

1 Like

2 posts were split to a new topic: How many observations are needed before a taxon makes it into the Computer Vision training package?

Kind of a name problem. Asparagus setaceus is Common Asparagus Fern. Asparagus officinalis is Wild Asparagus, although it is also the cultivated species. I think the “Common” helps Asparagus setaceus get mapped everywhere (though it’s frost sensitive). The “Wild” may encourage people to think A. officinalis can’t be the name for what they recognize as the cultivated plant even when it’s growing wild. Not sure there’s a solution, but I think these common names are part of the problem.

2 Likes

I’ve just added related topic:
https://forum.inaturalist.org/t/auto-ids-of-gastropods-are-completely-useless/16315
Though it’s not the case of few separate taxa, but rather most gastropods. Especially annoying and common wrong auto-suggestions are Helicina orbiculata, Monacha cantiana, Helicella itala, Mesodon thyroidus, Arianta arbustorum, Theba pisana, Praticolella, but if to fix them then simply something else wrong will show up.

1 Like

(I am hurt) Perhaps the potted ones got the right name because they were, actively cultivated, the way people use the word outside iNat.
But yes I can well imagine that gone feral it would be a nasty weed.

1 Like

Trachyandra is currently blooming around Cape Town.
iNat is determined to offer various northern hemisphere options, that I have never heard of. They then show up on the distribution map as lonely onlies.

Bog aspodel? Not.

I wasn’t referring to the wild Asparagus in South Africa. You know far more about your truly wild and native specimens than I do, of course.

I speak of the miss-applied computer-vision suggestions all over the world. It seems that the ones I found in the “needs ID pool,” that is, not marked as captive even if they were captive, were more often grossly incorrect—yarrow, fennel, dogfennel, horsetails, various conifers, rosemary, the list goes on—than the ones already marked captive. Those marked captive were more likely to be the real deal or at least something else in the genus, although still with plenty of mistakes. I can brainstorm a few reasons why this might be, including

  • new users who rely a lot on computer vision also tend to rely on other people to mark their captive plants captive, so a lot of the captive pool has had a second set of eyes on it already
  • some people who keep houseplants do know what their plants are called
  • wild asparagus diversity is huge, but there’s only really 4ish species that are very popular in cultivation, so if the person was certain the word “asparagus” was in the common name somewhere, the pool of species to choose from is smaller.
2 Likes

Not worth trying to fix, but I’ve run into it a couple times now. The computer vision labels piles of rocks as “elephant seal”. Not a bad choice, really.

12 Likes

The majority of Usnea longissima photos are other species of Usnea. It’s so bad I’m sure the computer vision can’t be accurately trained.

Of course, there are other species labeled Usnea longissima, mostly Alectoria species, but also Ramalina menziesii, Ramalina farinacea, Bryoria, and others lichens, as well as the moss Isothecium stoloniferum.

4 Likes

According to the Hawaii flora checklist, the invasive taxon here is Asparagus plumosus but was previously misidentified as A. setaceus for a long time. But now I see that iNat is proposing to merge them, based on POTW.

I’m operating on the assumption that they are the same/going to be merged. If the powers that be change their minds, I’m going to be in trouble :sweat_smile:

2 Likes

Hi, I’d like to suggest koyamaki, Sciadopitys verticillata. It’s a very unique-looking tree - relatively uncommon, but you occasionally see it cultivated as an ornamental. However, in my experience the AI never identifies a bona fide example as this species. It always suggests plants in the Pinaceae (often Deodar cedar, Cedrus deodara).

Here’s an example of what I’m talking about: https://www.inaturalist.org/observations/50834319
https://www.inaturalist.org/observations/40904746

2 Likes

Just putting this out there, but Ceratostigma seems to never be identified as such by CV, it often goes for Hydrolea.

It looks like oxalismtp has gone through all of the Hydrolea misidentifications, which is good.

thanks!

2 Likes

Catharanthus roseus (Madagascar Periwinkle) can have white flowers with a red center, instead of the more common all-pink flowers. The computer vision is labeling these plants as Cerbera manghas (Grey Milkwood), which is a tree and has narrower leaves.

I have added Iresine diffusa to the list. Computer Vision often identifies as this anything with brightly colored leaves. This leads to all sorts of misidentifications, especially in fall.

https://www.inaturalist.org/observations/identify?quality_grade=any&ident_taxon_id=130870&without_taxon_id=130870 is a good search to find all the ones already corrected, which might need a nudge or two to get them on the right track.

I’m getting anxiety from seeing so many planted plants not marked, please, as observers often have no idea about it - mark them as cultivated.

2 Likes

I just added three entries and they got deleted - can we discuss things before they get deleted instead of just being deleted with no other indication?